PRIOR
ART

42

0

180+

July felt like the month the IP world moved from debating whether AI changes patent practice to arguing about how. Japan tightened its inventive step standard for AI-assisted applications. The USPTO’s foreign representation rule took effect today: a structural shift that arrives without fanfare but reshapes thousands of workflows overnight. And the world’s five largest patent offices are now formally aligned on AI examination, with a dedicated working group established. The question “where does automation end and accountability begin” is no longer hypothetical. This issue takes a position.

Japan just raised the bar for AI-assisted patent applications

The JPO’s Status Report 2026 (March 2026) tightened the inventive step standard for AI-assisted inventions. Examiners are now directed to determine whether a claimed invention reflects a genuine technical contribution beyond what a person skilled in the art, aided by commonly available AI tools, could independently achieve. The March 2026 examination handbook revision added new case examples covering generative AI applications and AI estimation methods. Voluntary disclosure of AI involvement in prosecution is now encouraged; a mandatory requirement is expected to follow.

Applications drafted without explicitly articulating human-directed design choices are now more exposed at JPO. If your team is prosecuting AI-related applications in Japan (or planning to), the specification language needs to be reviewed against the new standard before your next office action lands. Pendency is 9.1 months on standard examination. You won’t get a second chance to reframe before scrutiny arrives.

Read the JPO Status Report 2026 →


Foreign applicants without U.S. counsel are now out of compliance

The USPTO’s final rule requiring all foreign-domiciled patent applicants and patent owners to be represented by a registered U.S. practitioner took effect July 20, 2026. The rule applies to new filings, pending applications, and existing patent matters. Approximately 97% of foreign-origin non-provisionals already list a U.S. practitioner, but the remaining 3%, plus any applicant whose U.S. counsel has changed since the application was filed, face immediate exposure.

This is not a future risk. It is a today risk. Any foreign-domiciled applicant or patent owner without a listed U.S. practitioner of record is currently non-compliant. Review your portfolio for any affected matters, particularly applications filed directly, PCT nationals without U.S. counsel, or matters where previous counsel has changed.

Read the final rule details →


The world’s five largest patent offices have formally established a shared AI working group

At the 19th IP5 Heads of Office Meeting on June 12, 2026, hosted by the JPO in Tokyo, patent chiefs from Japan, Europe, South Korea, China, and the United States reviewed progress on the IP5 NET/AI Roadmap and agreed on new directions for AI cooperation. The five offices formally established a dedicated AI working group tasked with advancing practical discussions and coordinated initiatives across the IP5 system. The day prior, the offices convened with industry representatives to exchange views on AI’s current status and prospective use in examination.

Five offices accounting for the majority of global patent filings are now formally structured around AI coordination, not aspirationally but with a working group mandate and a signed joint statement. For IP teams prosecuting across these jurisdictions, the question is no longer whether AI examination standards will converge, but at what pace. The working group’s output will define what “aligned AI examination” actually means in practice, and it will shape prosecution strategy across every major market.

Read the JPO IP5 meeting summary →

EPO

EPO Guidelines 2026 entered into force April 1, integrating Enlarged Board of Appeal decisions G 1/23 and G 1/24, and introducing a new section holding applicants responsible for the accuracy of AI-assisted submissions, even when the AI got it wrong. epo.org →

USPTO

The USPTO cut the threshold for requiring additional information in unintentional-delay revival petitions from 2 years down to 1 year, effective June 24, 2026. For any practice with applications that missed prosecution deadlines in that 1–2 year window, revival petitions just got harder. federalregister.gov →

IFI CLAIMS

Agentic AI now accounts for 15% of all U.S. AI patent filings, up from 7% two years ago. Globally, agentic applications surged 59% over the same period. Nvidia leads all filers with 225 global agentic applications in 2025; OpenAI filed 35 patents worldwide; Anthropic filed 8. ificlaims.com →

One debate. Every issue. No easy answers.


Modern neural machine translation has reached quality thresholds that make full review unnecessary for routine, lower-stakes applications. For high-volume portfolios filing in multiple languages, mandatory review of every translation creates bottlenecks and cost overhead that slow the entire prosecution timeline. Applied selectively (with AI review confined to mature translation pairs and technically straightforward descriptions), the risk is manageable and the efficiency gains are real.

Modern neural machine translation has reached quality thresholds that make full review unnecessary for routine, lower-stakes applications. For high-volume portfolios filing in multiple languages, mandatory review of every translation creates bottlenecks and cost overhead that slow the entire prosecution timeline. Applied selectively (with AI review confined to mature translation pairs and technically straightforward descriptions), the risk is manageable and the efficiency gains are real.

Both sides agree AI translation quality has improved. The disagreement is about accountability. When an AI-generated translation is filed and later found to have shifted claim scope, who owns that outcome: the AI vendor, the filing firm, or the client? Until that question has a clear contractual and legal answer (and right now, it doesn’t), the risk calculus favours human review. Every time.


AI disclosure requirements expanding beyond JPO

EPO Guidelines 2026 entered into force April 1, integrating Enlarged Board of Appeal decisions G 1/23 and G 1/24, and introducing a new section holding applicants responsible for the accuracy of AI-assisted submissions, even when the AI got it wrong. epo.org →

EP validation strategy under the UPC: the data is starting to arrive

Eighteen months into the Unified Patent Court’s operation, early data on whether IP teams are actually changing their EP validation decisions is emerging. Initial signals suggest more selective validation, not broader coverage, driven by the UPC’s cost and uncertainty profile. We’re watching whether that pattern holds as more renewal and opposition decisions come through.

Machine translation accountability: the first high-profile case

As AI-generated translations become more common in patent prosecution, questions around liability for translation-induced claim narrowing are moving from theoretical to practical. The field is waiting for the first high-profile enforcement case where an AI translation error demonstrably altered claim scope. When it arrives, it will change how every firm approaches the review question in The Gray Area above.

across global PCT and EP translation and filing services

in 25+ years of PCT and EP translation and filing work

to get accurate PCT cost breakdowns in Legal Studio

209,518

That’s 23% generative AI and 9% agentic AI. Agentic alone has grown 59% over the past two years. The volume of AI-related prosecution is no longer something to monitor. It’s a workflow challenge to solve.


Park IP 20th Annual Scotch Tasting

Thursday, October 27, 2026 · Del Frisco’s Double Eagle Steakhouse, Washington DC

AIPLA Annual Meeting 2026

October 29–31 · The Westin Washington DC

PRIOR ART